Developmental Genetics of Corolla Tube Formation: Role of the tasiRNA-ARF Pathway and a Conceptual Model

We show that the tasiRNA-ARF pathway is required for corolla tube formation in monkeyflowers (Mimulus) and propose a new conceptual model highlighting the central role of auxin-directed synchronized growth in congenital petal fusion.

Abstract

Over 80,000 angiosperm species produce flowers with petals fused into a corolla tube. The corolla tube contributes to the tremendous diversity of flower morphology and plays a critical role in plant reproduction, yet it remains one of the least understood plant structures from a developmental genetics perspective. Through mutant analyses and transgenic experiments, we show that the tasiRNA-ARF pathway is required for corolla tube formation in the monkeyflower species Mimulus lewisii. Loss-of-function mutations in the M. lewisii orthologs of ARGONAUTE7 and SUPPRESSOR OF GENE SILENCING3 cause a dramatic decrease in abundance of TAS3-derived small RNAs and a moderate upregulation of AUXIN RESPONSE FACTOR3 (ARF3) and ARF4, which lead to inhibition of lateral expansion of the bases of petal primordia and complete arrest of the upward growth of the interprimordial regions, resulting in unfused corollas. Using the DR5 auxin-responsive promoter, we discovered that auxin signaling is continuous along the petal primordium base and the interprimordial region during the critical stage of corolla tube formation in the wild type, similar to the spatial pattern of MlARF4 expression. Auxin response is much weaker and more restricted in the mutant. Furthermore, exogenous application of a polar auxin transport inhibitor to wild-type floral apices disrupted petal fusion. Together, these results suggest a new conceptual model highlighting the central role of auxin-directed synchronized growth of the petal primordium base and the interprimordial region in corolla tube formation.

INTRODUCTION

About one-third of the ∼275,000 angiosperm species produce flowers with petals fused into a corolla tube (i.e., sympetalous), forming a protective enclosure around nectaries and reproductive organs. Corolla tubes have evolved multiple times independently across the angiosperm tree of life (Endress, 2011), most notably in the common ancestor of the Asterids, a clade containing more than 80,000 species (Schonenberger and Von Balthazar, 2013). Subsequent elaboration in length, width, and curvature has led to a great variety in corolla tube shapes that enabled asterid species to exploit many specialized pollinator groups (e.g., bee flies, hawkmoths, hummingbirds, nectar bats), which in turn drives rapid plant speciation (Muchhala, 2006; Hermann and Kuhlemeier, 2011; Paudel et al., 2015; Lagomarsino et al., 2016). As such, the corolla tube has long been considered a key morphological innovation that contributed to the radiation of the Asterids (Endress, 2011). Despite its critical importance in the reproductive success and evolution of such a large number of species, the corolla tube remains one of the least understood plant structures from a developmental genetics perspective (Specht and Howarth, 2015; Zhong and Preston, 2015; Rebocho et al., 2017; Preston et al., 2019).

Historically, corolla tube formation has been the subject of extensive morphological and anatomical studies (Boke, 1948; Kaplan, 1968; Govil, 1972; Nishino, 1976, 1978, 1983a, 1983b; Erbar, 1991; Erbar and Leins, 1996; Kajita and Nishino, 2009; El Ottra et al., 2013). In particular, numerous studies have described the detailed ontogenetic process of corolla tube development in one subgroup of the asterid clade, the Lamiids, which contains some classical plant genetic model systems such as snapdragon (Antirrhinum), petunia (Petunia), and morning glory (Ipomoea; Govil, 1972; Nishino, 1976, 1978, 1983a, 1983b; Singh and Jain, 1979; Erbar, 1991; Vincent and Coen, 2004; Kajita and Nishino, 2009; Erbar and Leins, 2011). A common theme emerging from these studies is that during the early stage of petal development, petal primordia are initiated separately, followed by rapid extension of the petal bases toward the interprimordial regions, which also grow coordinately, causing congenital fusion of the petal primordia and formation of the corolla tube. Little is known, however, about the genetic control of this early-phase lateral extension of the petal base or the nature of this coordinated interprimordial growth.

A handful of genes have been implicated in corolla tube formation. Loss-of-function mutations in FEATHERED (FE) in Japanese morning glory (Ipomoea nil) and MAEWEST (MAW) in petunia (Petunia × hybrida), both generated by transposon insertions, result in unfused corollas (Iwasaki and Nitasaka, 2006; Vandenbussche et al., 2009). FE and MAW encode KANADI and WOX transcription factors, and their Arabidopsis (Arabidopsis thaliana) orthologs are KANADI1 and WOX1, respectively. In addition, ectopic accumulation of Arabidopsis TEOSINTE BRANCHED1, CYCLOIDEA, AND PCF TRANSCRIPTION FACTOR5 (TCP5) fused with a repressor motif in Ipomoea disrupted corolla tube formation (Ono et al., 2012). However, whether the endogenous TCP5 ortholog in Ipomoea is involved in corolla tube development is unclear. It was also reported that transient knockdown of the petunia genes encoding the NAC transcription factors NO APICAL MERISTEM (NAM) and NAM-like protein16 (NH16) by virus-induced gene silencing (VIGS) caused decreased petal fusion (Zhong et al., 2016). However, the interpretation of this result was confounded by the observation that occasional flowers produced on escape shoots of the loss-of-function nam mutants had normal corolla tubes (Souer et al., 1996). In addition, NAM expression was not detected between petal boundaries in petunia (Preston et al., 2019). The fact that these genes were characterized from different plant systems and through different methods (transposon insertion alleles, heterologous expression of a chimeric repressor, and VIGS) makes it challenging to interpret their genetic relationships and their precise functional roles in corolla tube formation.

One solution around this problem is to systematically analyze corolla tube mutants in a single model system. To this end, we employ a new genetic model system, the monkeyflower species Mimulus lewisii, mainly for its relative ease of use in chemical mutagenesis and Agrobacterium (Agrobacterium tumefaciens)-mediated in planta transformation (Owen and Bradshaw, 2011; Yuan et al., 2013b; Yuan, 2019). M. lewisii is a typical bumblebee-pollinated species with a conspicuous corolla tube (Figure 1A). Through EMS mutagenesis, we have generated a dozen recessive mutants (named flayed) with unfused corollas. Here, we report the characterization of one group of mutants, caused by loss-of-function mutations in two genes that are required for the biogenesis of trans-acting short interfering RNAs (tasiRNAs).

Figure 1.

Phenotypic Characterization of the flayed Mutants.

(A) Face and side views of the wild-type M. lewisii (inbred line LF10) corolla. d, dorsal; l, lateral; v, ventral.

(B) Side view of flayed1 corolla.

(C) and (D) Face views of flayed2 (C) and the flayed1 flayed2 double mutant (D).

(E) Pistil of the wild type and flayed2. The pistil phenotype of flayed1 and the double mutant is the same as that of flayed2.

(F) Quantitative comparison of petal width in the wild type (n = 18), flayed1 (n = 10), flayed2 (n = 12), and the double mutant (n = 12). Each flower was sampled from a distinct plant. Detailed measurement data are presented in Supplemental Table 1.

(G) Overall shape of the fourth leaf (the largest leaf) of mature plants.

(H) Quantitative comparison of length and width of the fourth leaf, with the same sample sizes as in (F). The leaf samples came from the same set of plants as in (F).

(I) Abaxial epidermal cells of dorsal petal lobes.

(J) Width of five contiguous abaxial epidermal cells of the dorsal petal lobes in the wild type (n = 15) and flayed2 (n = 15).

Error bars in (F), (H), and (J) represent 1 sd. Asterisks indicate differences from the wild type (**, P < 0.01, Student’s t test), and the lack of asterisks indicates no significant difference from the wild type. Bars in (A) to (E) = 5 mm.

Among the tasiRNA loci characterized to date, TRANS-ACTING SIRNA3 (TAS3) is the most widely conserved, found in virtually all land plants (Xia et al., 2017). TAS3 transcripts bear two binding sites for microRNA390 (miR390), which triggers the production of phased tasiRNAs, including the highly conserved tasiARF that targets AUXIN RESPONSE FACTOR2 (ARF2), ARF3, and ARF4 transcripts (Allen et al., 2005; Williams et al., 2005; Axtell et al., 2006). The tasiRNA-ARF3/4 regulatory module plays a critical role in leaf adaxial/abaxial polarity and blade expansion (i.e., lamina growth) in both eudicots (Fahlgren et al., 2006; Garcia et al., 2006; Hunter et al., 2006; Yan et al., 2010; Yifhar et al., 2012; Zhou et al., 2013) and monocots (Nagasaki et al., 2007; Nogueira et al., 2007; Douglas et al., 2010). Consistent with previous studies, here we demonstrate that in the M. lewisii mutants, TAS3-derived tasiRNAs decrease dramatically in abundance and that MlARF3 and MlARF4 expression are upregulated. Importantly, we show that malfunction of the tasiRNA-ARF pathway in M. lewisii mutants impedes the early lateral expansion of the petal primordium bases and their coordinated interprimordial growth, most likely through a change in auxin homeostasis, leading to unfused petal primordia. Integrating our molecular and phenotypic analyses of the tasiRNA-ARF pathway and the spatial patterns of MlARF4 expression and auxin responses in M. lewisii with historical insights from morphological and anatomical studies of various sympetalous species, we propose a new conceptual model for the genetic control of corolla tube formation. This model offers logical connections among the previous sporadic reports of corolla tube mutants and makes clear predictions that can be readily tested using the Mimulus system.

RESULTS

Phenotypic Characterization of the flayed1 and flayed2 Mutants

The recessive mutants flayed1 to flayed3 recovered from EMS mutagenesis using the inbred line LF10 were morphologically indistinguishable. Pair-wise crosses suggested that they belong to two complementation groups, flayed1 and flayed2 (flayed3 being allelic to flayed2; Figures 1B and 1C). In addition to having unfused petals, these mutants displayed carpel fusion defects, with phenotypes varying from flower to flower within the same plant. Most mutant flowers had two fused carpels, as in the wild type, but with partially split stigmas with more than two lobes (Figure 1E), while some flowers had two almost completely separate styles. The length of mutant pistils was also reduced compared with the wild type (Figure 1E). No obvious phenotypes were observed in the stamens of these mutants.

Another notable feature of the flayed1/2 mutants was the reduced width of their lateral organs. The dorsal and lateral petals showed ∼30% decrease in width compared with the wild type and ∼37% decrease for the ventral petal (Figure 1F; Supplemental Table 1; and see the Supplemental Data Set for summary of statistic tests). The mutations also substantially reduced leaf width by ∼40%, but length remained unaffected (Figures 1G and 1H; Supplemental Table 1). To determine whether the reduction in petal width was due to changes in cell number, cell size, or both, we measured the width of abaxial (outer) epidermal cells of the dorsal petal lobe for both wild-type and flayed2 flowers. Because petal lobe abaxial epidermal cells are irregularly shaped (Figure 1I), we performed width measurements on five contiguous cells to account for the variation among individual cells within the same sample. We detected no significant differences in cell width between the wild type and flayed2 (Figure 1J), suggesting that the difference in petal width between the mutant and the wild type is primarily due to a difference in cell number (i.e., number of cell divisions).

Unlike the morning glory fe mutant (Iwasaki and Nitasaka, 2006) or the petunia maw mutant (Vandenbussche et al., 2009), flayed1/2 mutants did not show any defects in tissue adaxial/abaxial polarity. Instead, the flayed1/2 mutants closely resembled the petunia mutant choripetala suzanne (chsu), which also exhibits unfused corollas, variable carpel fusion defects, and narrower leaves with normal adaxial/abaxial polarity. Unfortunately, the molecular identity of CHSU is still unknown.

FLAYED1 and FLAYED2 Are the Orthologs of Arabidopsis AGO7 and SGS3, Respectively

To identify the causal genes behind the flayed1 and flayed2 mutants, we analyzed each mutant using a genomic approach that combines the advantages of bulk-segregant analysis and comparison of single-nucleotide polymorphism (SNP) profiles between multiple EMS mutants (see Methods), as demonstrated in a previous study by LaFountain et al. (2017). We narrowed down the causal mutations for flayed1 and flayed2 to 38 and 19 candidate SNPs, respectively (Supplemental Tables 2 and 3). The vast majority of these SNPs located in noncoding, repetitive sequences, with only two or three mutations resulting in amino acid changes in each mutant (Supplemental Tables 2 and 3). Notably, each mutant showed a single point mutation leading to a premature stop codon, in the ortholog of Arabidopsis ARGONAUTE7 (AGO7) for flayed1 and in SUPPRESSOR OF GENE SILENCING3 (SGS3) for flayed2 (Figures 2A and 2B; Supplemental Tables 2 and 3). AGO7 and SGS3 are part of the same tasiRNA biogenesis pathway (Peragine et al., 2004; Yoshikawa et al., 2005; Chen, 2010), which would explain the identical phenotypes exhibited by flayed1 and flayed2. Furthermore, sequencing the coding sequence (CDS) of MlSGS3 in flayed3, which is allelic to flayed2, revealed an independent mutation that also led to a premature stop codon (Figure 2B). Together, these results suggested that MlAGO7 and MlSGS3 were the most promising candidate genes for FLAYED1 and FLAYED2.

Figure 2.

FLAYED1 and FLAYED2 Are the Orthologs of Arabidopsis AGO7 and SGS3.

(A) and (B) Schematics of MlAGO7 (A) and MlSGS3 (B) gene structure, with causal mutations indicated. Black boxes, exon; lines, intron. Scale bars = 100 bp.

(C) and (D) Flower phenotypes of 35Spro:MlAGO7 (C) and 35Spro:MlSGS3 (D) transgenics in the flayed1 and flayed2 mutant backgrounds, respectively (top, face view; bottom, side view). The proportions of fully rescued, partially rescued, and nonrescued lines are shown in parentheses.

(E) and (F) Relative transcript levels of MlAGO7 (E) and MlSGS3 (F) in 5-mm floral buds of four representative, fully rescued overexpression lines compared with the corresponding mutant backgrounds, as determined by RT-qPCR. MlUBC was used as the reference gene. R1 and R2 represent two individual plants. Error bars represent 1 sd from three technical replicates. Asterisks indicate differences from the corresponding mutants (*, P < 0.05 and **, P < 0.01, Student’s t test).

To validate these gene identities, we introduced the full-length CDS for MlAGO7 and MlSGS3 into the flayed1 and flayed2 mutant background, respectively, driven by the cauliflower mosaic virus 35S promoter. Among 29 independent 35Spro:MlAGO7 lines in the flayed1 background, 13 showed a fully rescued phenotype that was indistinguishable from the wild type; four lines showed partial rescue, with petal and leaf width similar to that of the wild type but with unfused petals (Figure 2C). Similarly, six of the 18 MlSGS3 overexpression lines in the flayed2 background displayed a fully rescued phenotype and two exhibited partial rescue (Figure 2D). RT-qPCR assessment of MlAGO7 and MlSGS3 expression in 5-mm floral buds showed that expression levels of the transgenes were approximately fourfold to 64-fold higher than those of the corresponding endogenous genes in the fully rescued lines (Figures 2E and 2F). These results confirmed that MlAGO7 and MlSGS3 were indeed the causal genes underlying flayed1 and flayed2.

Knowing the causal genes and mutations also allowed direct genotyping of a flayed1 × flayed2 F2 population to identify flayed1 flayed2 double mutants, which were phenotypically indistinguishable from the single mutants (Figures 1D and 1F to 1H). This result further indicated that MlAGO7 and MlSGS3 functioned in the same genetic pathway in Mimulus, as expected.

The flayed1/2 Phenotypes Are Primarily Mediated Through the tasiRNA-ARF Pathway

AGO7 and SGS3 are necessary components of the miR390-TAS3-ARF pathway (Figure 3A), and the highly conserved, TAS3-derived tasiARFs are known to play a critical role in leaf polarity and lamina growth by repressing ARF3/4 expression. We therefore hypothesized that the flayed1/2 phenotypes (e.g., reduced width of lateral organs) were primarily mediated through the tasiRNA-ARF pathway. Three clear predictions derive from this hypothesis: (1) the abundance of TAS3-derived small RNAs, including tasiARFs, should be much lower in the mutants compared with the wild type; (2) the M. lewisii ARF3/4 orthologs should be upregulated in the mutants; and (3) artificial upregulation of the M. lewisii ARF3/4 orthologs in the wild-type background should recapitulate the flayed1/2 phenotypes. We chose to focus on ARF3/4 rather than ARF2, the other known target of tasiARFs (Williams et al., 2005; Marin et al., 2010), because most of the traits controlled by ARF2 in Arabidopsis (e.g., seed size, leaf senescence, floral organ abscission, fruit ripening, and dehiscence; Ellis et al., 2005; Okushima et al., 2005; Schruff et al., 2006; Lim et al., 2010) were unaffected in our flayed1/2 mutants compared with the wild type.

Figure 3.

Small RNA Analysis.

(A) Schematic of the miR390-TAS3-ARF pathway. The orange and red lines represent the two tandem tasiARFs (see Supplemental Figure 1 for detailed annotations).

(B) Small RNA counts in the wild type, flayed1, and flayed2. Two biological replicates, indicated by 'a' and 'b' in the parentheses, were sequenced for each genotype. Each biological replicate consisted of pooled 5-mm floral buds from a distinct plant.

To test the first prediction, we sequenced the total small RNA pool from young floral buds (5 mm) from the wild type and the flayed1 and flayed2 mutants. Like most other angiosperms, M. lewisii had two kinds of TAS3 genes (each represented by only a single copy in the M. lewisii genome): TAS3S contained a single, centrally located tasiARF, whereas TAS3L contained two tandem tasiARFs (Supplemental Figure 1; Xia et al., 2017). We detected no TAS3S-derived small RNAs in any of the sequenced samples, suggesting that the TAS3S gene was not expressed. By contrast, we detected TAS3L-derived small RNAs to the level of ∼600 per million reads in the wild type, but these decreased >50-fold in both flayed1 and flayed2 (Figure 3B). In particular, tasiARFs were almost entirely absent from the mutant samples (Figure 3B). These results validated the first prediction.

To test the second prediction, we first searched the M. lewisii genome for ARF3/4 homologs and identified a single putative ortholog for each gene. Similar to ARF3/4 in other species, both MlARF3 and MlARF4 had two binding sites with sequences complementary to tasiARF (Figures 4A and 4B). RT-qPCR measurements in 5-mm floral buds showed that MlARF3 was upregulated by ∼1.7- to 2.5-fold and MlARF4 was upregulated by ∼2.7- to 3.7-fold in the single and double mutants (Figure 4C). This moderate upregulation of ARF3/ARF4 in the ago7 and sgs3 mutant backgrounds was very similar to previous reports in Arabidopsis (Garcia et al., 2006; Hunter et al., 2006), supporting the role of tasiARF in fine-tuning ARF3/4 expression levels.

Figure 4.

The flayed1/2 Phenotypes Are Primarily Mediated through Upregulation of MlARF3/4.

(A) and (B) Schematics of MlARF3 (A) and MlARF4 (B) gene structure. Red bars represent tasiARF binding sites. The nucleotides highlighted in red are synonymous substitutions at the two tasiRNA binding sites that were introduced in the 35Spro:mMlARF3 and 35Spro:mMlARF4 constructs to circumvent tasiRNA repression. Scale bars = 100 bp.

(C) Relative transcript levels of MlARF3 and MlARF4 in 5-mm floral buds of mutant plants compared with the wild type, as determined by RT-qPCR. MlUBC was used as the reference gene. Error bars represent 1 sd from three biological replicates (i.e., three individual plants) for each genotype.

(D) to (F) Leaf and flower phenotypes of the strongest 35Spro:mMlARF3 (D), 35Spro:mMlARF4 (E), and double transgenic line (F). Left, the wild type; right, the transgenic line. The red arrowheads indicate points of petal separation.

(G) to (I) Relative transcript levels of MlARF3 in the 35Spro:mMlARF3 line with the strongest phenotype (G), MlARF4 in the 35Spro:mMlARF4 line with the strongest phenotype (H), and MlARF4 in two 35Spro:mMlARF4 transgenic lines without any obvious phenotypes (I). MlUBC was used as the reference gene. Error bars represent 1 sd from three biological replicates (5-mm floral buds from three different plants for the wild type and from three lateral branches of the same T1 plant for the transgenic lines).

Asterisks in the bar graphs in (C) and (G) to (I) indicate differences from the wild type (*, P < 0.05 and **, P < 0.01, Student’s t test), and the lack of asterisks indicates no statistically significant difference from the wild type.

To test the third prediction, we transformed the wild type with a tasiARF-insensitive version of MlARF3 (mMlARF3) and MlARF4 (mMlARF4) with several synonymous substitutions in the tasiARF binding sites (Figures 4A and 4B), driven by the 35S promoter. We obtained 7 independent 35Spro:mMlARF3 and 14 35Spro:mMlARF4 lines. In each case, only two transgenic lines showed obvious phenotypes: their leaves were very similar to the flayed1/2 mutants (i.e., narrower than the wild type) with partially split corollas (indicated by the red arrowheads in Figures 4D and 4E). RT-qPCR experiments on 5-mm floral buds of the transgenic lines with even the strongest phenotypes showed only moderate overexpression of MlARF3/4 relative to the wild type (approximately twofold to fourfold; Figures 4G and 4H). Examination of two random 35Spro:mMlARF4 lines without obvious phenotypes showed no increase in MlARF4 expression levels (Figure 4I). The lack of 35Spro:mMlARF3/4 lines with strong transgene expression contrasted with the ectopic expression of MlAGO7 and MlSGS3 (Figures 2E and 2F) as well as that of pigment-related transcription factors in M. lewisii (Yuan et al., 2014; Sagawa et al., 2016; Ding et al., 2020), where the same 35S promoter readily drove transgene expression levels >10-fold higher than the endogenous genes. One possible explanation for this observation is that transgenic lines with very strong ARF3/4 expression in M. lewisii are seedling-lethal, as implicated by similar experiments in tomato (Solanum lycopersicum; Yifhar et al., 2012). Nevertheless, our results showed that a moderate upregulation of MlARF3/4 indeed fully recapitulated the leaf phenotype and partially recapitulated the flower phenotype of the flayed1/2 mutants. Furthermore, a double transgenic line derived from a cross between the strongest 35Spro:mMlARF3 and 35Spro:mMlARF4 lines showed dramatic petal fusion defects (Figure 4F), indicating that MlARF3 and MlARF4 may act synergistically in regulating corolla tube formation. Taken together, our results from transgenic manipulation of the MlARF3/4 expression levels suggested that the flayed1/2 phenotypes (narrow leaf and unfused corollas) were primarily mediated by the upregulation of MlARF3/4.

The tasiRNA-ARF Pathway Is Required for Preferential Lateral Expansion of the Bases of Petal Primordia and Coordinated Growth of Interprimordial Regions

To understand how malfunction of the tasiRNA-ARF pathway affects corolla tube formation in M. lewisii, we studied floral organogenesis in the wild type and the flayed2 mutant using scanning electron microscopy. Like other species in the lamiid clade (e.g., snapdragon, petunia, morning glory), M. lewisii petals initiated as five separate primordia (Figure 5A). Petal development lagged behind stamens in the early stages (Figures 5B and 5C), but by the time the corolla reached 0.5 mm in diameter (Figure 5D), petal development progressed rapidly and soon the stamens were enclosed in the corolla (Figures 5E to 5H). The developmental stage from 0.3 to 0.4 mm (corolla diameter) was critical for corolla tube formation: during this stage, the bases of the petal primordia quickly expanded laterally (to a conspicuously greater extent than the upper portion of the petal primordia; Figure 5M), and the interprimordial regions also grew coordinately, connecting the initially separate petal primordia. Floral organogenesis of flayed2 was very similar to that of the wild type during the early stages (before the corolla reaches 0.3 mm in diameter; Figure 5I). However, during the critical period (0.3 to 0.4 mm), we observed no preferential lateral expansion at the bases of the petal primordia in the mutant, manifested as the truncate shape of the petal primordium base (Figure 5N), in contrast to the semicircle shape seen in the wild type (Figure 5M). Notably, growth of the interprimordial regions was also arrested in flayed2, leading to a gap between two adjacent petal primordia (Figures 5J to 5L and indicated by the asterisk in Figure 5N).

Figure 5.

Scanning Electron Micrographs of M. lewisii Floral Development.

(A) to (H) Wild-type LF10. The developmental stages are marked on the bottom right of each image by the diameter (in mm) of the corolla. From 0.4 mm onward, sepals were removed to show the petals.

(I) to (L) flayed2.

(M) and (N) Detailed views of two adjacent petal primordia and the interprimordial region in the wild type (0.4-mm stage [M]) and flayed2 (0.5-mm stage [N]). Arrows indicate growth directions, and the asterisk in (N) marks the arrested interprimordial region.

p, petal; pi, pistil; st, stamen. Bars = 100 μm.

Given that disruption of tasiRNA biogenesis and the consequent upregulation of ARF3/4 have been shown to cause reduced lamina growth of lateral organs in multiple plant species (Peragine et al., 2004; Douglas et al., 2010; Yan et al., 2010; Yifhar et al., 2012; Zhou et al., 2013), it was perhaps not surprising to observe reduced lateral expansion at the bases of petal primordia in flayed2 compared with the wild type (Figures 5M and 5N). But how does this observation explain the arrest of upward growth of the interprimordial regions?

In a series of careful anatomical studies of various taxa in the Asterids, Nishino (1976, 1978, 1983a, 1983b) recognized that the cooperation between the marginal meristem activities at the base of the petal primordia and the upward growth of the interprimordial regions played a pivotal role in corolla tube formation, although the nature of this cooperation has been unclear. Considering this earlier insight, together with our own results (Figures 5M and 5N), we speculated that in the wild type, there was a molecular signal with continuous distribution in the marginal meristematic cells along the petal primordium base and the interprimordial region, stimulating synchronized growth between the two regions. In the flayed2 mutant, this growth signal may be reduced or disrupted in its spatial distribution, leading to arrested growth in the zone encompassing the margins of the primordium base and the interprimordial region. An obvious candidate for this putative signal would be the phytohormone auxin, which is known to promote localized tissue outgrowth while also suppressing the expression of organ boundary genes such as CUP-SHAPED COTYLEDON1 (CUC1) and CUC2 in Arabidopsis (Bilsborough et al., 2011).

Spatial Distribution of Auxin Response in Developing Corolla Buds Corresponds to Petal Growth Patterns in the Wild Type, and Auxin Homeostasis Is Altered in flayed2

To examine the spatial pattern of auxin responses in developing corollas, we introduced the auxin response reporter construct DR5rev:mRFPer (Gallavotti et al., 2008), driving the expression of RFP by the synthetic auxin-responsive promoter DR5, into the wild-type LF10. In the very early stage when petal primordia have just initiated (corresponding to a stage between Figures 5A and 5B), we detected strong RFP signal in both petal primordia and interprimordial regions, with a clear gap between the two (Figure 6A). As the corolla reached 0.4 mm in diameter, RFP signal became concentrated at the apex of the petal primordium and the “synchronized growth zone,” encompassing the margins along the primordium base and the interprimordial region (between the arrowheads in Figure 6B), with relatively weaker signal along the upper part of the petal primordium margin. It is worth noting that the spatial pattern of auxin response at this stage corresponded almost perfectly with the petal growth pattern leading to corolla tube formation (i.e., preferential lateral expansion of the petal primordium base and the coordinated upward growth of the interprimordial region; Figure 5M). As the corolla reached 0.5 mm in diameter, auxin responses remained continuous at the base of the petal primordium and the interprimordial region (demarcated by the arrowheads in Figure 6C), but the gap devoid of auxin response between the petal apex and the synchronized growth zone became more conspicuous than at the 0.4-mm stage. When the corolla reached 0.6 mm in diameter, RFP signal started to spread more evenly along the entire petal margin (Figure 6D), likely corresponding to later growth of the entire corolla. While all confocal images shown in Figure 6 were taken using DR5rev:mRFPer line 2, examination of several additional independent transgenic lines showed very similar results.

Figure 6.

Patterns of Auxin Response in the Developing Corolla Buds of the Wild Type and flayed2, as Reflected by the DR5rev:mRFPer Reporter.

(A) to (D) The wild type. The developmental stages are marked on the bottom right of each image by the diameter (in mm) of the corolla. The white arrowheads in (B) and (C) demarcate the synchronized growth zone encompassing the marginal meristematic cells at the base of the petal primordia and the interprimordial cells. ip, interprimordial region; p, petal; pi, pistil; se, sepal; st, stamen. Bars = 100 μm.

(E) and (F) flayed2, shown in the same style as the wild type. The overall fluorescence signal is much weaker in the mutant compared with the wild type. Petal primordia and interprimordial regions are outlined in white dashed lines. White arrows mark the gap at the junction between the petal primordium base and the interprimordial region.

To test whether auxin homeostasis was altered in the flayed2 mutant, we crossed DR5rev:mRFPer line 2 with flayed2 and analyzed F2 individuals homozygous for the flayed2 mutation and with the DR5rev:mRFPer transgene. When imaged under the same microscopy parameters, flayed2 corollas showed much weaker RFP signal overall than the wild type (Figures 6E and 6F). In particular, the junction between the petal primordium base and the interprimordial region (arrows in Figures 6E and 6F) showed no RFP signal at all, consistent with the lack of synchronized growth between these two regions in the mutant (Figure 5N). The relatively weak RFP signal in the mutant pistil (Figures 6E and 6F) also correlated with the split stigma/style phenotype (Figure 1E).

Exogenous Application of Polar Auxin Transport Inhibitor Disrupts Corolla Tube Formation

If auxin was indeed the signaling molecule that stimulated synchronized growth between the petal primordium base and the interprimordial region, exogenous application of polar auxin transport inhibitors during the early stages of corolla development may disrupt the spatial pattern of auxin signaling and thereby interfere with corolla tube formation. To test this prediction, we treated wild-type lateral shoot apices with 1-N-naphthylphthalamic acid (NPA) using a “spray/vacuum infiltration” approach (see Methods for details). About 30% of the ∼250 flowers that opened 3 to 4 weeks after the NPA treatment showed various defects in corolla tube formation (Figure 7A), whereas none of the ∼250 flowers produced by mock-treated plants showed any obvious phenotypic alterations. Reciprocally, exogenous application of the auxin indole-3-acetic acid (IAA) to flayed2 lateral shoot apices resulted in the rescue of corolla tube formation in 5 of the ∼250 flowers (Figure 7B), whereas none of the flowers on the mock-treated plants showed any phenotypic rescue. This relatively low frequency (2%) of corolla tube rescue by IAA treatment in the mutant, compared with the moderate frequency (30%) of corolla tube disruption by NPA treatment in the wild type, was not unexpected, as it is much easier to destroy the delicate spatial pattern of auxin signaling than to restore it by exogenous applications of the corresponding chemicals.

Figure 7.

NPA and IAA Treatments.

(A) Exogenous application of 5 μM NPA to the apices of lateral shoots of wild-type M. lewisii resulted in various corolla tube defects in ∼30% of the flowers.

(B) Exogenous application of 5 mM IAA to flayed2 resulted in rescue of the corolla tube in 2% of the flowers.

MlARF4 Expression Coincides with the Spatial Pattern of Auxin Response in the Developing Corolla

To further investigate how the tasiRNA-ARF pathway may relate to the spatial patterning of auxin signaling in the developing corolla, we studied the spatial expression pattern of MlARF4. We focused on MlARF4 because it was functionally redundant with MlARF3 in corolla tube formation (Figures 4D to 4F), its transcript levels were higher than those of MlARF3 in wild-type floral buds based on unpublished transcriptome data, and its expression levels were upregulated to a higher degree than those of MlARF3 in the flayed1/2 mutants (Figure 4C). We transformed the wild-type LF10 with the reporter construct MlARF4pro:MlARF4-RFP, which contains an ∼8-kb MlARF4 genomic DNA fragment (∼3-kb promoter sequence upstream of the translation initiation site plus all exons and introns), with RFP cloned at the 3′ end. A careful examination of individual petals in multiple MlARF4pro:MlARF4-RFP transgenic lines showed that MlARF4 predominantly localized to the abaxial side of the developing petal (Figures 8A and 8B), resembling the ARF3/4 expression pattern in leaf primordia of Arabidopsis and tomato (Yifhar et al., 2012; Guan et al., 2017). More importantly, the MlARF4 expression pattern in the young corollas was very similar to the spatial pattern of auxin response. When the corolla was 0.5 mm in diameter, MlARF4 was concentrated in the synchronized growth zone (i.e., the interprimordial region and the adjacent primordium bases; Figure 8C), similar to the pattern of auxin response at the same developmental stage (Figure 6C). In 0.6-mm corollas, MlARF4 expression became more widespread along the petal margin (Figure 8D), reminiscent of the DR5:RFP signals observed earlier (Figure 6D).

Figure 8.

MlARF4 Expression Pattern in Individual Petals and Flower Buds of Wild-Type M. lewisii, as Revealed by the MlARF4pro:MlARF4-RFP Reporter.

(A) and (B) Confocal laser scanning microscopy of a single petal at the 3.0-mm corolla developmental stage. Representative orthogonal views are shown for the Z-stack of an individual petal of MlARF4pro:MlARF4-RFP line 10 (A) and MlARF4pro:MlARF4-RFP line 7 (B). Shown in the bottom green box and the right purple box are single orthogonal sections of cross-sectional planes in XZ and YZ, respectively. The white arrows point to the abaxial side of the petal. Note that the cross-sectional XZ plane in (B) captured the petal margin (indicated by the arrowheads), where MlARF4 is observed in both sides. Ab, abaxial; Ad, adaxial.

(C) and (D) MlARF4 expression in whole flower buds. ip, interprimordial region; p, petal; pi, pistil; st, stamen. Bars (bottom right corner of each panel) = 50 μm.

This coincidence of the MlARF4 expression pattern and DR5:RFP signal in the synchronized growth zone provided a plausible explanation for the reduced auxin response in the flayed2 mutant. MlARF3/4 can repress auxin responses through at least two nonexclusive mechanisms: (1) ARF3/4 are known to repress auxin signaling by competing with ARF activators (e.g., ARF5/6/7/8) for the auxin response element binding sites in the promoters of auxin-responsive genes (Tiwari et al., 2003; Vernoux et al., 2011; Lavy et al., 2016); (2) ARF3 has been shown to directly repress auxin biosynthesis and transport in Arabidopsis (Simonini et al., 2017). Therefore, upregulation of MlARF3/4 in the flayed2 mutant would be expected to exert stronger repression of auxin signaling in the synchronized growth zone, leading to arrested growth and lack of corolla tube formation.

It should be noted that there was also an obvious difference between the MlARF4pro:MlARF4-RFP signal and the DR5:RFP signal: the former was not found at the apex of the petal primordium, although the latter was (Figures 6C and 8C). This difference indicated that MlARF3/4 were not the only regulators of auxin signaling in the developing corolla.

DISCUSSION

This study represents the first step in a systematic effort toward understanding the developmental genetics of corolla tube formation using the new model system M. lewisii. We showed here that the tasiRNA-ARF pathway is required for synchronized growth between the petal primordium base and the interprimordial region at early stages of corolla development. Furthermore, we discovered that auxin signaling is continuous in this synchronized growth zone in the wild type, similar to the spatial pattern of MlARF4 expression. Upregulation of MlARF3/4 in the flayed2 mutant correlated with much reduced auxin response in the synchronized growth zone, explaining the inhibition of lateral expansion at the base of the petal primordia and complete arrest of the upward growth of the interprimordial regions observed in the mutant. Together, these results suggest a new conceptual model for the developmental genetic control of corolla tube formation, highlighting the central role of auxin signaling in the synchronized growth zone.

Role of the tasiRNA-ARF Pathway in Petal Fusion

Although this is the first detailed study investigating the role of the tasiRNA-ARF pathway in corolla tube formation, a previous study on leaf development in the family Solanaceae has mentioned in passing that malfunction of the tasiRNA-ARF pathway led to unfused corollas in tomato and tobacco (Nicotiana tabacum) flowers (Yifhar et al., 2012). Like Mimulus, the family Solanaceae belongs to the asterid clade. This suggests that the role of the tasiRNA-ARF pathway in corolla tube formation is likely conserved across asterid plants. More surprisingly, two other studies of leaf development in barrelclover (Medicago truncatula) and trefoil (Lotus japonicus), both belonging to the legume family (Fabaceae), also implicated an indispensable role for the tasiRNA-ARF pathway in petal fusion (Yan et al., 2010; Zhou et al., 2013). Typical legume flowers have separate dorsal and lateral petals, but the two ventral petals often fuse into a prow-shaped structure (i.e., the “keel”). In the ago7 mutants of M. truncatula and L. japonicus, the two ventral petals become separated. The family Fabaceae belongs to the clade Rosids. Given that the vast majority of rosid species (e.g., Arabidopsis) produce flowers with completely separate petals (i.e., polypetalous), with only a few exceptions in derived lineages (Zhong and Preston, 2015), it is clear that the tasiRNA-ARF pathway was independently recruited to enable petal fusion in the legume species.

There are two significant insights emerging from our molecular and phenotypic analyses of the tasiRNA-ARF pathway in Mimulus that were not known from the aforementioned studies that focused on leaf development: (1) the tasiRNA-ARF pathway is required not only for the lamina growth of petal primordia but also for the upward growth of the interprimordial regions (Figures 5M and 5N). In fact, we think that the synchronized growth between the petal primordium base and the interprimordial region is the key to corolla tube formation; (2) there is a tight correlation between the spatial pattern of MlARF4 expression, auxin response, and the growth of the interprimordial region in developing corollas (Figures 5M, 6B, 6C, and 8C), establishing a mechanistic link between the tasiRNA-ARF pathway, auxin signaling, and corolla tube formation.

The difference in spatial patterns of auxin response between flayed2 and the wild type (Figure 6) suggested that the role of the tasiRNA-ARF pathway in corolla tube formation may lie in the regulation of auxin homeostasis. Indeed, the coincidence of the MlARF4 expression pattern and the spatial pattern of auxin response suggests that MlARF3/4 play an important role in modulating the level of auxin signaling in the synchronized growth zone. In Arabidopsis, loss-of-function arf3 mutants displayed increased DR5:GFP signal in the gynoecium (Simonini et al., 2017). Conversely, upregulation of ARF3/4 homologs in the moss species Physcomitrium patens, either due to loss of SGS3 function or transgenic overexpression, resulted in decreased auxin response (Lavy et al., 2016; Plavskin et al., 2016). These observations suggest that ARF3/4 and their homologs can clearly function as negative regulators of auxin response, at least in some developmental contexts. Indeed, recent studies have shown that these repressive ARFs can negatively regulate auxin response by competing with activating ARFs for the same auxin response element binding sites in the promoters of auxin-responsive genes (Lavy et al., 2016) or directly repressing the transcription of auxin biosynthesis and transport genes via histone deacetylation (Simonini et al., 2017; Chung et al., 2019; Kuhn et al., 2020). Similar to the synchronized growth zone in the developing M. lewisii corolla, overlaps between the spatial pattern of ARF3/4 expression and auxin response also have been shown in lateral root primordia (Marin et al., 2010), founder cells of the reproductive primordia (Vernoux et al., 2011; Chung et al., 2019), and the abaxial side of leaf primordia (Guan et al., 2017) in Arabidopsis. These patterns suggest that in most developmental contexts, ARF3/4 do not act as a simple “off” switch to auxin response. Instead, they are usually part of a more complex genetic regulatory network with negative feedback loops that fine-tune local auxin responses (Marin et al., 2010; Vernoux et al., 2011; Plavskin et al., 2016). Since tasiARF small RNAs are a key component of these negative feedback loops (Marin et al., 2010; Plavskin et al., 2016), their complete depletion in the flayed2 mutant (Figure 3B) is expected to disrupt the robustness of the feedback regulation and explains the reduced level of auxin response in flayed2 (Figures 6E and 6F). The specific configuration of the auxin response genetic regulatory network in the developing corollas of sympetalous species like M. lewisii warrants further investigation.

Relationship between Petal Lamina Growth and Corolla Tube Formation

Given the well-known function of the tasiRNA-ARF pathway in lamina growth of lateral organs (Fahlgren et al., 2006; Garcia et al., 2006; Hunter et al., 2006; Nagasaki et al., 2007; Nogueira et al., 2007; Douglas et al., 2010; Yan et al., 2010; Yifhar et al., 2012; Zhou et al., 2013) and the reduced petal width of the flayed1/2 mutants, it is easy to overinterpret the significance of lamina growth of the petal primordia in the “fusion” of adjacent petals. However, the following observations suggest that lamina growth of the petal primordia per se is not the key to corolla tube formation. (1) In some of 35Spro:MlAGO7 and 35Spro:MlSGS3 transgenic lines (in the corresponding mutant background), petal width was restored to wild-type levels but the petals remained unfused (Figures 2C and 2D). In addition, we have several other yet-to-be-characterized flayed mutants with unfused corollas but normal petal width. These observations suggest that petal lamina growth at the wild-type level is not sufficient for tube formation. (2) Petal width of a previously characterized M. lewisii mutant, the dominant-negative actin mutant act1-D, is very similar to that of flayed1/2 (Ding et al., 2017), but the corolla tube of act1-D is intact. This suggests that reduced petal lamina growth does not necessarily affect petal fusion. Instead, we think it is the synchronization between the lateral expansion of the petal primordium base and the upward growth of the interprimordial region, most likely directed by auxin signaling, that holds the key to corolla tube formation.

A New Conceptual Model for the Developmental Genetic Control of Corolla Tube Formation

Two recent attempts to build a conceptual framework for floral organ fusion in general (Specht and Howarth, 2015) or petal fusion in particular (Zhong and Preston, 2015) both emphasized the genetic regulatory network underlying organ boundary formation and maintenance. The rationale for such emphasis was explicitly stated by Specht and Howarth (2015): “fusion as a process may more accurately be defined as a lack of organ separation.” While these attempts represent an important step toward a mechanistic understanding of the developmental process of corolla tube formation, to some degree they have neglected the insight provided by earlier morphological and anatomical studies (i.e., the importance of the cooperation between the rapid lateral expansion of the petal primordium base and the upward growth of the interprimordial region) and have not provided a logical explanation to the previously reported corolla tube mutants (e.g., morning glory fe and petunia maw; Iwasaki and Nitasaka, 2006; Vandenbussche et al., 2009) in terms of organ boundary formation or maintenance. Overemphasis on the “lack of organ separation” may represent an underestimation of the complexity of the corolla tube formation process, with the implication that simple loss-of-function mutations in some organ boundary genes in polypetalous species (e.g., Arabidopsis) could produce a functional corolla tube. Despite extensive mutagenesis of Arabidopsis in the past 40 years, such a mutant has not been reported.

Integrating our results on the tasiRNA-ARF pathway and the spatial pattern of auxin signaling in M. lewisii with previous reports of corolla tube mutants (Iwasaki and Nitasaka, 2006; Vandenbussche et al., 2009) as well as historical insights from anatomical studies (Nishino, 1978, 1983a, 1983b), we propose a new conceptual model for the developmental genetic control of corolla tube formation (Figure 9). At the heart of the model is auxin-induced synchronized growth of the marginal meristematic cells at the base of the petal primordia and the interprimordial cells, providing a molecular explanation for the cooperation between the petal primordium base and the interprimordial region observed in anatomical studies. Upstream of this core module is the genetic regulatory network controlling adaxial/abaxial polarity and lamina growth of lateral organs (Nakata and Okada, 2013; Tsukaya, 2013; Kuhlemeier and Timmermans, 2016), which is conserved in a wide range of angiosperms and is somehow coopted in sympetalous species to regulate auxin homeostasis in the synchronized growth zone. Downstream of this core module lies the genetic network responsible for localized tissue growth induced by auxin signaling.

Figure 9.

A Conceptual Model for the Developmental Genetic Control of Corolla Tube Formation in M. lewisii.

(A) and (B) During phase I (A), auxin activity maxima promote the initiation of individual petal primordia. During this phase, there is also a broad zone of auxin response in the interprimordial region, which leads to the initiation of upward growth around the entire circumference of the petal whorl and formation of the corolla tube during phase II (B). Blue and orange color indicate the abaxial and adaxial sides of the petal, respectively. Yellow color indicates the interprimordial zone.

(C) Genetic relationships among the three hypothetical modules. The delicate spatial pattern of auxin signaling plays a central role in directing the synchronized growth between the petal primordium base and the interprimordial zone and forms the core module of the model. The dynamics of auxin response in the developing corolla is at least partly regulated by the leaf adaxial/abaxial polarity network characterized in Arabidopsis. The downstream effector genes responsible for the auxin-induced localized tissue growth in the synchronized growth zone remain to be explored. GRN, genetic regulatory network.

This model can readily explain the phenotypes of loss-of-function mutations in morning glory FE and petunia MAW, which encode KANADI and WOX transcription factors, respectively (Iwasaki and Nitasaka, 2006; Vandenbussche et al., 2009). Together with the tasiRNA-ARF pathway, these transcription factors are part of the genetic network regulating leaf adaxial/abaxial polarity and lamina growth (Nakata and Okada, 2013; Tsukaya, 2013; Kuhlemeier and Timmermans, 2016). Recent studies in other model systems, such as Arabidopsis, barrelclover, and tobacco, have demonstrated that these polarity regulators largely function by modulating auxin homeostasis (Tadege et al., 2011; Huang et al., 2014; Simonini et al., 2017). According to our model, interfering with this genetic regulatory network is expected to reduce or disrupt the continuous auxin response in the synchronized growth zone, as shown in the flayed2 mutant (Figures 5N, 6E, and 6F), consequently resulting in unfused corolla. Our model also provides a plausible explanation for the petal fusion defects observed when a chimeric repressor of Arabidopsis TCP5 was overexpressed in the Japanese morning glory (Ono et al., 2012). Chimeric repressors of CIN-like TCP transcription factors (e.g., Arabidopsis TCP3/4/5) are known to decrease AGO7 and TAS3 expression in Arabidopsis (Koyama et al., 2010), which would be expected to upregulate ARF3/4 and inhibit corolla tube formation in sympetalous species.

One unresolved issue is whether organ boundary genes, such as the homologs of CUC/NAM, play an important role in corolla tube formation, as previously proposed (Specht and Howarth, 2015; Zhong and Preston, 2015). A recent observation in snapdragon (Rebocho et al., 2017) appears to support this proposition: the expression of the CUC/NAM ortholog, CUPULIFORMIS (CUP), is cleared from the interprimordial regions shortly after petal initiation but later is reactivated in the sinuses between adjacent corolla lobes. Computational modeling suggested that this “gap” of CUP expression (between the base of the corolla and the sinuses) is necessary for corolla tube formation (Rebocho et al., 2017), presumably by releasing the repression of localized tissue growth in the synchronized growth zone. However, if the role of CUC-like genes is to repress tissue growth at the petal-petal junction, downregulation or loss of function of NAM/CUP should increase the degree of petal fusion. Instead, VIGS knockdown of NAM and its paralog NH16 in the sympetalous petunia caused decreased petal fusion (Zhong et al., 2016). Furthermore, in situ hybridizations of NAM in petunia flowers detected no expression at all in petal-petal boundaries, arguing against a critical role of NAM in defining petal-petal boundaries in petunia (Preston et al., 2019). The exact cause of these conflicting observations is unclear; perhaps future experiments involving ectopic expression of CUC-like genes in sympetalous flowers will shed light on this problem.

Given the central role of auxin-directed synchronized tissue growth in corolla tube formation, our model predicts that, in addition to the leaf polarity/lamina growth network, disturbance of other genetic pathways that regulate auxin biosynthesis, polar transport, and signaling within and between petal primordia, in transgenic or mutant plants of sympetalous species, could also result in defective corolla tubes. The availability of multiple corolla tube mutants, the ease of bulk-segregant analysis to identify mutant genes, and the amenability of Agrobacterium-mediated in planta transformation make Mimulus a favorable system to test these predictions and to dissect the detailed molecular mechanisms and developmental process of corolla tube formation.

METHODS

Plant Materials and Growth Conditions

We performed EMS mutagenesis on the Mimulus lewisii inbred line LF10, as described by Owen and Bradshaw, (2011). We used the inbred line SL9 to generate flayed F2 mapping populations. We grew all plants on FAFARD soil mix #2 (Sun Gro Horticulture) in the University of Connecticut research greenhouses, under natural light supplemented with sodium vapor lamps, ensuring a 16-h daylength and a light intensity of 110 to 160 μmol m⁻² s⁻¹.

Phenotypic Characterization

To quantify phenotypic differences between the mutants and the wild type, we measured the widths of the dorsal, ventral, and lateral petals using a digital caliper. We also measured the lengths and widths of the fourth leaf (the largest leaf) of mature plants. To further evaluate whether the width difference is due to changes in cell number, cell size, or both, we measured the width of the abaxial epidermal cells of the dorsal petal lobe following a procedure previously described by Ding et al. (2017).

Genomic Analyses for Causal Gene Identification

To identify the causal genes underlying flayed1 and flayed2, we employed a hybrid strategy that combines the advantages of bulk-segregant analysis and genome comparisons between multiple EMS mutants, as previously described by LaFountain et al. (2017). Briefly, we produced an F2 population for each mutant by crossing the homozygous mutant (generated in the LF10 background) to the polymorphic inbred line SL9. We pooled DNA samples from 96 F2 segregants displaying the mutant phenotype (i.e., homozygous for the causal mutation) with equal representation. We then prepared a small-insert library for the pooled sample before sequencing on an Illumina HiSeq 2000 platform at the University of North Carolina High Throughput Sequencing Facility. We generated ∼213 and 448 million 100-bp paired-end reads for flayed1 and flayed2, respectively.

We mapped the short reads to the LF10 genome assembly version 1.8 (http://monkeyflower.uconn.edu/resources/) using CLC Genomics Workbench 7.0 (Qiagen). The causal mutation should be (1) homozygous in the pooled sample (i.e., 100% SNP frequency in the “F2 reads-LF10 genome” alignment) and (2) unique to each mutant (i.e., any shared 100% SNPs between mutants are most likely due to assembly error in the reference genome, nonspecific mapping of repetitive sequences, or private SNPs in our laboratory LF10 stock). After comparisons with the SNP profiles of previously published mutants, guideless (Yuan et al., 2013a), rcp1 (Sagawa et al., 2016), act1-D (Ding et al., 2017), and rcp2 (Stanley et al., 2020), we narrowed the causal mutations to 39 and 19 candidate SNPs for flayed1 and flayed2, respectively (Supplemental Figure 2; Supplemental Tables 2 and 3).

Small RNA Sequencing and Analyses

For small RNA sequencing, we extracted total RNA using the Spectrum Plant Total RNA Kit (Sigma-Aldrich) from 5-mm floral buds of LF10, flayed1, and flayed2 (two biological replicates for each genotype). We then constructed small RNA libraries using Illumina TruSeq Small RNA Sample Preparation Kits, with 1 μg of total RNA as starting material. We sequenced the libraries on an Illumina HiSeq 2500 platform at the Delaware Biotechnology Institute. Small RNA reads were quality-controlled and adaptor-trimmed before calculating tasiRNA abundance, as described by Xia et al. (2017).

RT-qPCR

We performed RNA extraction and cDNA synthesis as previously described by Yuan et al. (2013b). We diluted cDNA samples 10-fold before RT-qPCR. All RT-qPCRs were performed using iQ SYBR Green Supermix (Bio-Rad) and a CFX96 Touch Real-Time PCR Detection System (Bio-Rad). Samples were amplified for 40 cycles of 95°C for 15 s and 60°C for 30 s. Amplification efficiencies for each primer pair were determined using critical threshold values obtained from a dilution series (1:4, 1:8, 1:16, 1:32) of pooled cDNAs. The M. lewisii ortholog of the Arabidopsis (Arabidopsis thaliana) ubiquitin-conjugating enzyme gene At5g25760/UBC21, MlUBC, was used as a reference gene, as described by Yuan et al., (2013b). Relative expression of each target gene compared with the reference gene was calculated using the formula (E_ref)^CP(ref)/(E_target)^CP(target). Primers used for RT-qPCR are listed in Supplemental Table 4.

Plasmid Construction and Plant Transformation

To generate the 35Spro:MlAGO7 and 35Spro:MlSGS3 constructs for rescue experiments, we PCR-amplified the full-length CDS of MlAGO7 and MlSGS3 from wild-type LF10 cDNAs using the Phusion enzyme (New England Biolabs). For each gene, we cloned the amplified fragment into the pENTR/D-TOPO vector (Invitrogen) before using it as a template to amplify a linear fragment containing the CDS flanked by the attL1 and attL2 sites using M13 primers. We then recombined this linear fragment into the Gateway vector pEarleyGate 100 (Earley et al., 2006), which drives transgene expression from the cauliflower mosaic virus 35S promoter. To generate the 35Spro:mMlARF3/4 constructs, we synthesized the CDS of sRNA-insensitive forms for MlARF3 (mMlARF3) and MlARF4 (mMlARF4) with synonymous substitutions in the two tasiRNA recognition sites at GenScript and before cloning into the pEarleyGate 100 destination vector, as described for the 35Spro:MlAGO7 and 35Spro:MlSGS3 constructs. To generate the MlARF4pro:MlARF4-RFP reporter construct, we cloned an ∼8-kb genomic fragment for MlARF4, including the ∼3-kb promoter sequence upstream of the translation start site and all exons and introns, into the pDONOR207 vector before recombination into the pGWB653 destination vector (Nakamura et al., 2010). All plasmids were verified by sequencing before being transformed into Agrobacterium (Agrobacterium tumefaciens) strain GV3101 for subsequent plant transformation, as described by Yuan et al. (2013b). Primers used for plasmid construction and sequencing are listed in Supplemental Table 5.

Scanning Electron Microscopy

We fixed flower buds overnight in formalin-acetic-alcohol at 4°C, dehydrated for 30 min through a 50, 60, 70, 95, and 100% (v/v) ethanol series. Samples were then critical-point dried, mounted, and sputter-coated before observation using a NOVA NanoSEM with Oxford EDX at 35 kV at the University of Connecticut Bioscience Electron Microscopy Laboratory.

Confocal Microscopy

We observed early-developing floral buds (with sepals removed if necessary) carrying the DR5rev:mRFPer construct by laser confocal microscopy in the red channel as Z-stacks. All fluorescence images were acquired at the University of Connecticut Advanced Light Microscopy Facility using an A1R confocal laser scanning microscope (Nikon) equipped with a 20×/0.45 S Plan Fluor ELWD objective lens with the shoot apex temporarily under water. For the DR5rev:mRFPer reporter lines, all imaging was done using the same confocal settings, with a laser line of 558 nm, pinhole size of 46 μm, and laser power of 15. For the MlARF4pro:MlARF4-RFP reporter lines, imaging was done with a laser line of 558 nm, pinhole size of 56 μm, and laser power of 21.

NPA and IAA Treatments

To prepare plants for NPA and IAA treatments, we pinched the shoot apex of seedlings at the eight-leaf stage (i.e., four pairs of leaves) to promote lateral growth. Once each plant produced at least a dozen lateral shoots, we sprayed the apex of each lateral shoot with 5 μM NPA in 0.2% (v/v) DMSO (for the wild type) or 5 mM IAA in water (for the flayed2 mutants) using a mist bottle. We then placed the plants in a vacuum chamber and applied a pressure of 15 in Hg for 2 min before being released quickly. Each treatment was applied to five plants with corresponding mock treatment controls (i.e., 0.2% [v/v] DMSO solution or water was applied to wild-type and flayed2 plants, respectively).

Accession Numbers

Short-read data have been deposited in the National Center for Biotechnology Information Short Read Archive (BioProject PRJNA423263); small RNA data have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus (GSE108530); annotated gene sequences have been deposited in GenBank: MlAGO7 (MG669632), MlSGS3 (MG669633), MlARF4 (MG669634), and MlARF3 (MF084285); the MlUBC sequence is available from Mimubase (http://mimubase.org/) under accession number ML6G280000.

Supplemental Data

• Supplemental Figure 1. Annotation of the TAS3 transcripts in Mimulus lewisii.

• Supplemental Figure 2. SNP profile comparison between different EMS mutants.

• Supplemental Table 1. Phenotypic measurements of the wild-type, flayed1, flayed2, and the double mutant.

• Supplemental Table 2. flayed1 candidate SNPs.

• Supplemental Table 3. flayed2 candidate SNPs.

• Supplemental Table 4. Primers used for RT-qPCR.

• Supplemental Table 5. Primers used for plasmid construction and sequence verification.

• Supplemental Data Set. Summary of statistical tests.

DIVE Curated Terms

The following phenotypic, genotypic, and functional terms are of significance to the work described in this paper:

• AGO7 Gramene: AT1G69440

• AGO7 Araport: AT1G69440

• WOX1 Gramene: AT3G18010

• WOX1 Araport: AT3G18010

• NAM Gramene: AT1G52880

• NAM Araport: AT1G52880

• SGS3 Gramene: AT5G23570

• SGS3 Araport: AT5G23570

• ARF4 Gramene: AT5G60450

• ARF4 Araport: AT5G60450

• CUC2 Gramene: AT5G53950

• CUC2 Araport: AT5G53950

• ARF3 Gramene: AT2G33860

• ARF3 Araport: AT2G33860

• RCP1 Gramene: KR053165

• RCP1 Araport: KR053165